Research On Remaining Useful Life Prediction Of Rolling Bearings Based On LSSVR-HSMM

In actual engineering,condition-based maintenance(CBM)can effectively ensure equipment reliability,reduce maintenance risks and various costs,and has great practical significance.The remaining useful life(RUL)prediction is one of the key technologies for condition-based maintenance and has received extensive attention in recent years.Therefore,in order to solve the problem of remaining useful life prediction in practice and improve the accuracy,this thesis studies the remaining useful life prediction based on least squares support vector regression/hidden semi-Markov model with rolling bearings as the research object.The main research contents of this thesis are:In this thesis,firstly,a method for predicting the remaining useful life of rolling bearings based on least squares support vector regression/hidden semi-Markov model is proposed and the problem of remaining useful life prediction in both the complete data and the missing data is studied in detail.This method improves some of the shortcomings in the remaining useful life prediction method based on hidden semi-Markov model,and can predict the remaining useful life of the rolling bearings effectively.In the specific implementation process of rolling bearing remaining useful life prediction,three key technologies are proposed for different problems.Aiming at the sensor signals processing problem of rolling bearing,a vibration signal processing technique based on multi-domain feature and kernel principal component analysis is proposed.Aiming at the problem of degradation state recognition of rolling bearing,a degradation state recognition technique based on hidden semi-Markov model for segmentation of observation sequences is proposed.Aiming at the problem of rolling bearing observation sequence prediction,an observation sequence prediction technique based on whale swarm algorithm /least squares support vector regression is proposed.These three technologies are the key technologies of the method,and guarantee the effect of the method.Then,this thesis has carried out specific application and experimental verification of the research method in the IEEE PHM 2012 bearings data set.The results show that the three key techniques are feasible and reliable,and can effectively solve different problems in the remaining life prediction process.At the same time,in both cases of complete data and missing data,the method proposed in this thesis has a good predictive effect on the remaining useful life of rolling bearings.Finally,this thesis summarizes the work of the full text,and combines the content of this paper to look forward to the future research direction.